The Association Between Abdominal Muscle and Type II Diabetes Across Weight Categories in Diverse Post-Menopausal Women

Abstract

Context:

Despite the key role of muscle in glucose regulation, little is known about the association between muscle area and prevalence of metabolic disorders, or the role low muscle may play in normal weight metabolic obesity.

Objective:

The objective was to assess the independent associations between both abdominal muscle and fat depositions (measured by computed tomography) and the prevalence of type II diabetes, and to explore the modifying role of weight category.

Design:

We conducted a cross-sectional analysis of the 2001–2002 visit for the Rancho Bernardo Study, Filipino Women's Health Study, and Health Assessment Study of African American Women.

Setting and Participants:

Participants were 392 community-dwelling older women (mean age = 64) free of clinical cardiovascular disease.

Main Outcome Measure:

The main outcome was prevalence of type II diabetes, defined as use of anti-diabetes medication, fasting plasma glucose ≥ 126 mg/dL, and/or OGTT ≥ 200 mg/dL.

Results:

Adjusting for demographics, hypertension, estrogen use, lipids, smoking, physical activity, visceral fat area, and height, a greater muscle-to-total abdominal area ratio (MAR) was associated with lower odds of diabetes [OR = 0.63 per standard deviation, 95% CI (0.43–0.92), p = .02]. Higher visceral fat was associated with greater odds of diabetes in fully adjusted models including total muscle area [OR = 1.48, 95% CI (1.09, 2.01), p = .01]. Associations between MAR and diabetes were stronger for normal weight (BMI 18.5–24.9; OR = 0.32) than overweight/obese women (BMI ≥ 25, OR = 0.71, p-for-interaction = 0.046). Associations with visceral fat did not differ by BMI (p-for-interaction = 0.71).

Conclusions:

In older women, abdominal muscle area is inversely associated with type II diabetes independent of visceral adiposity, particularly for normal weight women.

While diabetes onset can lead to skeletal muscle loss, it is also possible that low muscle mass could be a diabetes risk factor. Glucose storage and consumption by lean muscle are key factors in glycemic regulation (1). In one study a higher ratio of estimated muscle-to-total body weight predicted lower prevalence of insulin resistance and prediabetes independent of central adiposity (2). However, muscle and adiposity in that study were estimated using standardized equations rather than directly measured equations (2). The association between diabetes and quantified muscle via computed tomography (CT) has not been reported.

To date, most studies assessing muscle area and metabolism have focused on overweight or obese individuals (3, 4). Diabetes is also prevalent in individuals classified as normal weight (BMI 18.5–24.9), though little is known about what contributes to metabolic dysfunction in these individuals. While some research suggests normal weight, dysglycemia may be due to higher distributions of visceral adipose tissue (VAT) (5), this explanation appears incomplete (6). It is plausible that, in the absence of excess adipose tissue, low muscle mass may play an especially important role in predisposition to insulin resistance and dysglycemia. However, no studies have reported the association between muscle and diabetes prevalence across weight strata.

Therefore, the purpose of the present study was to assess the association between abdominal muscle area measured by CT and type II diabetes in older women from the Rancho Bernardo Study, the University of California San Diego (UCSD) Filipino Women's Health Study, and the Health Assessment Study of African-American Women (HASAAW). We hypothesized that lower abdominal muscle area would be associated with higher prevalence of type II diabetes. We also assessed associations between diabetes and VAT to compare and adjust for the role of central adiposity in diabetes prevalence. Additionally, we explored the potential role of weight category in modifying the association between muscle area and diabetes.

Materials and Methods

Design and study participants

The Rancho Bernardo Study (RBS) is a prospective cohort study of community-dwelling older adults established between 1972 and 1974 in a suburb of San Diego, California. Since then, participants had periodic follow-ups. Additionally, Filipina women from the UCSD Filipino Women's Health Study and African-American women from the HASAAW cohorts were recruited in 1994–1999 using the same research protocol, staff, and diagnostic laboratories. Details of recruitment and inclusion/exclusion criteria for all three studies can be found elsewhere (7–10). The present study includes women from the RBS (n = 164), HASAAW (n = 124), and Filipino Women's Health Study (n = 104) who attended the 2001–2002 CT examinations and had complete abdominal muscle and fat, outcome (diabetes) and covariate data. Only women with no known heart disease were studied.

The RBS, UFWHS, and HASAAW studies were each approved by the Institutional Review Board of the University of California, San Diego. All participants gave written informed consent at all examinations.

Measurement of covariates

Detailed descriptions of covariate measurements, including demographics, lifestyle variables, lipids, and medical history, have been published previously (11). Weight and height were measured and used to calculate BMI (kg/m²), and participants were classified as normal weight (BMI = 18.5–24.9) or overweight/obese (≥ 25.0). No participants were underweight (BMI < 18.5).

Morning fasting blood samples were obtained by venipuncture after a 12-h fast at the 2001–2002 examination. A 75-g 2-h oral glucose tolerance test (OGTT) was administered after a minimum 8-h overnight fast at a 1992–1996 examination for Whites and the 1994–1999 examination for Filipina and African-American women. Fasting and 2-h postchallenge plasma glucose were measured using standard methods (11).

Prevalent diabetes definition

Prevalent diabetes was defined as antidiabetes medication use, fasting plasma glucose ≥ 126 mg/dL, or 2-h postchallenge glucose ≥ 200 mg/dL. Diabetes medication use, including insulin, was assessed by medication inventory. In order to take into account all available information, including the more sensitive OGTT measurements, prevalent diabetes was defined as a period prevalence, using the years 1992–2002. We also performed sensitivity analyses using only diabetes measures from the 2001–2002 visit.

Measurement of abdominal muscle and fat

Abdominal muscle and fat areas were measured from the 2001–2002 examination CT scans by three experienced CT analysts working on networked workstations running muscle segmentation software on the MIPAV platform (MIPAV Version 4.1.2, NIH). Labwide intra- and inter-rater reliability for CT scans ranged from 0.85 to 0.99.

A transverse slice at ∼L4-L5 disc space of 6 mm thickness was selected. VAT was measured using the innermost border of the abdominal muscle wall following the parietal peritoneum. Muscle segmentation was performed on the psoas, paraspinal, oblique, and rectus abdominus muscles (see Larsen et al for details) (11). Pixel attenuation between −190 to −30 Hounsfield Units (HU) was defined as fat and 0–100 HU as lean muscle. Voxels outside these ranges were undefined.

Total abdominal muscle was defined by summing over the left and right side for each muscle group. Muscle-to-abdominal area ratio (MAR) was obtained by dividing the total abdominal muscle area by the total abdominal area. VAT-to-total abdominal area was similarly computed.

Statistical analysis

Univariate associations of participant characteristics with prevalent diabetes status were assessed using t test, χ², or Wilcoxon tests as appropriate. Staged multivariate logistic regression was used to examine the associations of muscle and fat areas with prevalent diabetes in separate models, starting with age and race/ethnicity (model 1), plus ever smoking, exercise ≥ three times per week (12), hypertension, HDL cholesterol, LDL cholesterol, triglycerides, and estrogen use (model 2), plus body size and composition (height and VAT to muscle models and height and total abdominal muscle to fat models) in model 3. Abdominal muscle and fat, as well as ratios, were modeled per standard deviation.

Similarly staged multivariate linear regression models were used to determine the association of abdominal muscle and VAT areas with fasting and 2-h postchallenge glucose as continuous outcomes.

Interactions between BMI group and abdominal muscle/VAT areas were tested in models adjusted for age. Interactions were tested on a multiplicative scale in logistic regression models with prevalent diabetes as the outcome.

P values < .05 were considered statistically significant and SPSS 20 was used for all analyses.

Results

Participant characteristics overall and by prevalent diabetes are presented in Table 1. Mean ± SD age was 64 ± 7 years with an overall mean BMI of 27 kg/m². Race/ethnicity, BMI, ever smoking, systolic, and diastolic blood pressures, lipid levels, estrogen use, VAT, and total abdominal muscle differed significantly by diabetes status (P < .05).

Table 1.

Participant characteristics overall and by type II diabetes status^a

	Overall	No Type II Diabetes	Prevalent Type II Diabetes	P value
	n = 392	n = 326	n = 66
Age (y)	64 (7)	64 (7)	63 (6)	.67
Ethnic Group (%)
White	42	46	20	<.001
African-American	32	33	27
Filipina	26	21	53
BMI, kg/m2 (continuous)	27 (5)	26 (4)	28 (5)	.01
BMI category (%)
< 25	43	46	29	.027
25 to < 30	39	36	51
≥ 30	18	18	20
Ever smoker (%)	42	45	26	.004
Exercise ≥ 3×/week (%)	71	71	68	.434
Hypertension (%)	52	50	64	.040
Systolic BP, mmHg	132 (21)	131 (21)	137 (20)	.032
Diastolic BP, mmHg	77 (9)	76 (9)	79 (9)	.024
Hypertension Meds (%)	34	33	39	.13
HDL cholesterol, mg/dL	62 (17)	64 (17)	54 (134)	<.001
LDL cholesterol, mg/dL	127 (34)	124 (31)	140 (43)	.001
Triglycerides, mg/dL	122 (68)	116 (65)	156 (74)	<.001
Estrogen Use (%)	52	55	38	.01
Oral Glucose Tolerance Test	143 (65)	123 (31)	264 (83)	<.001
Visceral fat, cm2	107 (53)	102 (51)	135 (57)	<.001
Total abdominal muscle, cm2	84 (19)	85 (19)	78 (21)	.008

^aMean (sd) for continuous variables, and % for categorical or binary variables; P values by t-test, χ-square or Wilcoxon test as appropriate.

Total abdominal muscle, VAT, MAR, and VAT-to-total abdominal area ratio were significantly associated with prevalent diabetes in unadjusted models (Table 2). Associations remained significant after further adjustment for race/ethnicity, age, lifestyle factors, height, and either total muscle or VAT. Limiting the analysis to diabetes measured only at the 2001–2002 visit did not materially change the results.

Table 2.

Association of abdominal muscle, fat, and ratios to total abdominal area with prevalent type II diabetes^a

	Total Abdominal Muscle OR (95% CI); P	VAT OR (95% CI); P	MAR OR (95% CI); P	VAT/ Total Abdominal Area OR (95% CI); P
Unadjusted	0.70 (0.54, 0.92); 0.01	1.74 (1.36, 2.24); < 0.001	0.54 (0.40, 0.73); < 0.001	1.70 (1.31, 2.21); < 0.001
Model 1b	0.75 (0.56, 1.00); 0.05	1.64 (1.25, 2.14); < 0.001	0.54 (0.39, 0.75);< 0.001	1.56 (1.16, 2.11); 0.04
Model 2c	0.75 (0.55, 1.01); 0.06	1.50 (1.11, 2.03); 0.01	0.58 (0.41, 0.82); 0.002	1.36 (0.98, 1.89); 0.07
Model 3d	0.73 (0.54, 0.99); 0.04	1.48 (1.09, 2.01); 0.01	0.63 (0.43, 0.92); 0.02	1.44 (1.02, 2.02); 0.04

^aPer standard deviation increase for total abdominal muscle (19.19 cm²), VAT (53.22 cm²), MAR (0.049), and VAT to total abdominal area ratio (0.062).

^bModel 1: Age and ethnic group.

^cModel 2: Model 1 + ever smoker, exercise, hypertension, HDL cholesterol, LDL cholesterol, triglycerides, estrogen use.

^dModel 3: Model 2 + VAT and height for MAR and total abdominal muscle models; Model 2 + total abdominal muscle and height for VAT and VAT/total abdominal cavity area models

The association of MAR with prevalent diabetes differed for normal weight vs overweight/obese women, p-for-interaction = 0.046. Specifically, in age-adjusted analysis, among those with normal BMI, MAR was significantly inversely associated with prevalent diabetes [OR = 0.32, 95% CI (0.16, 0.62), P = .001], while the association was not significant for overweight/obese women [OR = 0.71, 95% CI (0.48, 1.06), P = .091]. Results were similar upon further adjustment for race/ethnicity, lifestyle factors, body size, lipids, and hypertension (results not shown). Similar results were found for total abdominal muscle (p-for-interaction = 0.073). The interaction of BMI with VAT was not significant (p-for-interaction = 0.71).

A linear regression showed a significant inverse association between MAR and fasting plasma glucose [B = −2.54 mg/dL per SD, 95% CI (−4.93, 0.23), P = .05] and 2-h postchallenge glucose [−8.0 mg/dL, 95% CI (−13.55, −2.44), P = .005] after adjusting for demographics, lifestyle, hypertension, lipids, and glucose-lowering medications. However, neither association was significant when further adjusting for VAT and height. Conversely, each SD of VAT was associated with greater fasting glucose (3.62 mg/dL, 95% CI (1.10, 6.14), P = .005) and greater 2-h glucose (10.77 mg/dL, 95% CI (4.90, 16.63), P < .001) after adjusting for all covariates, including total abdominal muscle and height.

Discussion

As hypothesized, total abdominal muscle area and MAR both showed strong inverse relationships with diabetes prevalence. After adjusting for demographics, lifestyle, CVD risk factors, and VAT, each SD increase was associated with 0.37-fold and 0.27-fold reduced odds of prevalent diabetes for MAR and total muscle, respectively (Table 2). VAT was associated with 1.48-fold greater odds of diabetes in fully adjusted models including total muscle. In addition to being independent of one another, VAT and muscle associations with diabetes were similar in magnitude (though inverse), suggesting that both increased VAT and decreased muscle are strong, independent risk factors for type II diabetes.

The BMI category modified the relationship between muscle and diabetes. Greater muscle was more protective for normal-weight women, who had less than half the odds (OR = 0.32) of diabetes per SD than overweight/obese women (OR = 0.71). This could suggest a role of low muscle in normal weight metabolic obesity, and could explain why past studies using only overweight/obese individuals found no association between muscle and metabolic indicators (3, 4).

The cross-sectional analysis cannot determine the direction of association, and it is possible that some individuals had a normal BMI because of weight loss due to diabetes-related muscle wasting (13, 14). However, wasting has typically been found with extremity muscles, and is less likely to affect muscles of locomotion and stabilization in the abdomen, such as those analyzed here. As mentioned previously, one study found an association between low estimated muscle mass and prediabetes, when muscle wasting is less likely to have occurred (2). Additionally, individuals who are normal weight at the time of diabetes onset actually have approximately double the mortality rate as overweight or obese individuals, (15) leading to speculation that low lean mass could both increase risk for diabetes and contribute to unique complications after its onset (16).

Greater muscle could increase both glucose storage and consumption, enhancing glucose disposal. Skeletal muscle is a main storage site for circulating glucose (1), and increasing muscle mass improves insulin sensitivity and prevents insulin resistance (17). Lean muscle contraction is also a primary mechanism of glucose consumption (18, 19), and increasing muscle mass has been shown to increase daily resting energy expenditure (20).

Several strengths and limitations warrant comment. Participants were older females free of cardiovascular disease at the time of the CT; thus results may not generalize to broader populations. There were also relatively few cases of diabetes (n = 66); thus, interactions in particular, should be interpreted with caution. Finally, CT measures of muscle and fat were limited to the abdomen. Strengths of the current study include an ethnically diverse sample and thorough measures, including CT and semiautomated software to quantify fat and muscle areas. Also, CT scans and covariate measures across all race/ethnicity groups were performed in the same location and read by the same technicians, allowing for reliable between-group comparisons.

To our knowledge, this is the first study to examine the association between prevalent diabetes and CT-measured abdominal muscle, or to do so across weight categories. Future studies are needed to examine the risk of incident diabetes in individuals with low muscle area, and to expand current findings to new populations.